Development of a new model for predicting EDM properties of Cu-TaC compact electrodes based on artificial neural network method

• Main Authors: Ndaliman, Mohammed Baba, Al Hazza, Muataz Hazza Faizi, Khan, Ahsan Ali, Yeakub Ali, Mohammad
• Format: Article
• Language: English
• Published: American-Eurasian Network for Scientific Information (AENSI Publisher) 2012
• Subjects: T Technology (General)
• Online Access: http://irep.iium.edu.my/55331/1/EDM%201%282012%29.pdf; http://irep.iium.edu.my/55331/; http://www.ajbasweb.com/old/ajbas/2012/December%202012/192-199.pdf

Electrical discharge machining (EDM) is one of the non-traditional machining processes normally used in manufacturing very hard materials that are electrically conductive. Tool electrodes form one of the main components of the machining system. The major properties that determine the suitability of such electrodes are electrical conductivity, thermal conductivity and density. The objective of this paper is to present the use of Artificial Neural Network (ANN) architecture in modeling these properties. In the research, Cu-TaC electrode compacts were produced at two levels each of the composition and the compacting pressures from copper and tantalum carbide powders for use in EDM. The compositions of the Cu-TaC are made of 30 % and 55 % wt of TaC, while the compacting pressures are 1, 500 psi and 3,000 psi. They were subjected to sintering at temperatures of 450°C and 850 °C. The properties were measured before and after sintering. Results showed that the sintered electrodes are not suitable for EDM because they lost their electrical conductivity. The presintered electrodes (green compacts) were however found to suitable for EDM. Artificial neural network technique with 16 experimental runs was used to develop the new models for predicting the electrical conductivity, thermal conductivity and density of the green compacted electrodes. The models were been built by using MATLAB 2009b. Results show that ANN models are capable of predicting the electrode properties with high degree of prediction accuracy compared to the experimental results